Application and optimization of reference change values for Delta Checks in clinical laboratory

Abstract Background Delta check is a patient‐based QC tool for detecting errors by comparing current and previous test results of patient. Reference change value (RCV) is adopted in guidelines as method for delta check, but the performance is not verified. We applied RCV‐based delta check method to patients' data and modified for application. Materials and methods Reference change value were calculated using results of internal QC materials and biological variation data. Test results of 17 analytes in inpatients, outpatients, and health examination recipients were collected. The detection rates of currently used delta check method and those of RCV‐based method were compared, and the methods were modified. Results Reference change value‐based method had higher detection rates compared to conventional method. Applied modifications reduced detection rates. Removing the pairs of results within reference interval reduced detection rates (0.42% ~ 10.92%). When RCV was divided by time interval, the detection rates were similar to prior rates in outpatients (0.19% ~ 1.34%). Using RCV multiplied by twice the upper limit of reference value as cutoff reduced the detection rate (0.07% ~ 1.58%). Conclusions Reference change value is a robust criterion for delta check and included in clinical laboratory practice guideline. However, RCV‐based method generates high detection rates which increase workload. It needs modification for use in clinical laboratories.

the development of automatic analyzers, and other technological advances, delta check has a low positive predictive value for specimen mix-ups. 3 Nowadays, delta check seems to be more useful to become aware of clinically significant changes in the patient status.
Commonly used delta check methods include delta difference, delta percent change, rate difference, and rate percent change.
Delta difference is defined as the absolute difference between current results and previous results. Delta percent change is defined as the percentage of the difference between the current results and the previous results divided by the previous one.
Rate difference is defined as the delta difference divided by the interval between the current and past testing time, and the rate percent change is defined as the delta percent change divided by the interval between the current and the previous testing time.
The selection of the delta check method depends on the extent of the test value, the distribution of the test value, etc. 4 There is no consensus regarding specific methods such as which analyte should be applied to the delta check and which cutoff should be set. Currently, each medical institution has different rules for delta check.
There are two methods to determine the limit of delta check, that is, using biological variation and using patient data. CLSI EP33 "Use of Delta Checks in the Medical Laboratory" provides a detailed description of how to use the Reference Change Value (RCV) as a method of delta check limit determination using biological variation. RCV is the minimum difference of a measurement from a reference value that is considered as distinguishable from measurement uncertainty, 5 and biological variation and analytical imprecision are used to calculate it. Measurement uncertainty is an expression of the statistical variance of a value due to a measured quantity. 6 No measurement method is exact and when a quantity is measured, the value depends on various factors, such as the measurement system, the measurement method, the proficiency level of the measurer, and the environment. Even if an amount is measured several times in the same manner and situation, the measured value will differ from one measurement to another. 7 An analytical imprecision can be obtained by reflecting the concept of uncertainty, and several attempts have been made to apply RCV calculated from this and the previously known biological variation data to delta check, 8,9 but which are not satisfactory yet. In this study, the delta check limit determined by using RCV is applied to the actual patients' data, and some modifications were tried toward the RCV-based delta check method.
Patient specimens and QC materials were tested using automated chemistry analyzers. A dedicated calibrator and a dedicated reagent provided by the manufacturer of each type of equipment were used. Roche Cobas C-8000 (Roche diagnostics) was used for inpatient and health screening recipient samples, and Beckman Coulter AU5800 (Beckman Coulter) was used for outpatient samples.

| Calculating RCV
Based on CLSI EP29-A, the standard deviation of the inspection values of the internal quality control material for 1 year was used to calculate the uncertainty using the top-down approach. 5

| Modifying the RCV-based delta check method
In the delta check rule setting process using RCV, some modification was performed in order to reduce the excessive detection rate. First, we excluded pairs of test results if both test results constituting the pair were within the reference interval of the analyte. The detection rate at this time was compared with the detection rate before the modification. Second, for analytes whose current delta check rule uses the rate difference and rate percentage change, the percent change was compared with the RCV with a 99% CI value divided by the interval of the test date, and the detection rate was calculated and compared with the detection rate before modification. As a third method, in the case of an analyte whose current delta check method uses the delta difference and rate difference, RCV with a 99% CI value multiplied by twice the upper normal limit value of the reference range was   compared with the absolute difference of each pair and the detection rate was calculated and compared with those before the modification.

| RCV calculation
Reference change value was calculated using CVa and CVi, and RCV with a 99% CI showed a significant difference from 4.4% to 136% depending on the analyte (Table 3). CVa did not vary significantly according to the analyte (1.05 ~ 7.42), but CVi showed a large fluctuation depending on the analyte, that is, 0.60 ~ 36.80.
For analytes with CVi smaller than 5.0%, such as calcium, protein, albumin, sodium, potassium, and chloride, the RCV was smaller than 20%. And for analytes with a CVi larger than 12.2%, such as AST, ALT, total bilirubin, and direct bilirubin, showed a large RCV of more than 50%.

| Application of the RCV-based delta check method to patients' data
In terms of the detection rate, both methods using 95% CI and 99% CI tend to have a higher detection rate compared to those using traditional cutoff (Table 4). This trend was more pronounced in outpatients than in inpatients. In exception, in patients with ALT and direct bilirubin, the detection rate was reduced when RCV was applied.

| Modification of the RCV-based delta check method
We constituting the pair were within the reference interval (Table 4). In this case, the detection rate was lower than 10% in the group of health examination recipients. The detection rate decreased by more than 50% in cholesterol and total bilirubin for inpatients and in cholesterol, total bilirubin, and potassium for outpatients.
Second, if the existing delta check method has a rate difference or rate percent change, the RCV is compared with the rate percent change to reflect the time factor for the analyte (Figure 1).
Applying this method, the detection rate in the outpatient group was 0.00 ~ 1.44%, similar to the conventional detection rate. In the inpatient group, except for ALP, all of them showed a reduced detection rate.
As the third modification method, when the current delta check method uses a delta difference or rate difference, the value obtained by multiplying the RCV value by twice the upper limit of the reference value was used as the delta check cutoff. As a result of this method, all of the analytes except for glucose showed much lower detection rate of less than 2% in both the inpatient and outpatient groups ( Figure 2).

| D ISCUSS I ON
The concept of the delta check was originally proposed by This uncertainty concept was introduced in the field of laboratory medicine, and the calculation of uncertainty and the concept of

RCV are introduced in CLSI EP29-A Expression of Measurement
Uncertainty in Laboratory Medicine. 5 The CLSI EP33 Use of Delta Checks in the Medical Laboratory also describes the use of RCVs for delta check cutoff settings. 15 Studies using RCV when setting up the delta check rule were recently published. 8,9 In our study, as the simple application of RCV without modification showed over-detection in all the analytes, it was considered that it would be difficult to apply RCV to delta check methods in clinical laboratories without modification.
Delta check is the best tool currently available for detecting specimen misidentification. 15 However, the introduction of the barcode system and the development of automated specimen processing can reduce the utility of the delta check in detecting the

(C) General health examination recipients
Analyte Detection rate (%) In calculating RCV, we first decided on the confidence interval.

RCV-based with 99% CI excluding within-RI pairs
In this study, RCV was calculated using 99% confidence interval.
The confidence interval of 95% for setting the delta check cutoff in the clinical laboratory tended to produce higher detection rate.
The over-detection tendency did not disappear even with the 99% confidence interval, suggesting that an additional modification was required for the RCV-based delta check method.
Among the additional modifications, there was a method to exclude the pair from the delta check alert if both values were within the reference interval. The rationale for coming up with this method is that the main purpose of the delta check is detecting the laboratory error, but it can be also used to detect the change in a patient's condition. This modification can increase the probability that the delta check alert has actual clinical significance, and this method has actually contributed to the reduction of over-detection in analytes such as cholesterol and total bilirubin. On the other hand, in the current delta check rule used by our laboratory, positive delta and negative delta are set differently in analyte such as creatinine, AST, ALT, ALP, total bilirubin, BUN, and direct bilirubin. In Ko's study, the positive delta is larger than the negative delta in the analyte that reflects tissue damage and in the analyte that can rise due to a patient's condition. 8 Considering this aspect, though it is not dealt with in this study, when RCV is applied, both limits are set equal, and it should be considered whether this is a proper delta check rule.
In glucose, the detection rate of 10% or less was not achieved in any of the delta check rules used in this study. This may be due to the clinical characteristics of the analyte. Glucose has a very large variation during a single day, depending on diet or blood sugar depressants or insulin administration. Therefore, it is difficult to regard a large change in glucose as a laboratory error such as specimen mix-up. Of the total 46 institutions in the Q-Probe study conducted by CAP, 16 used glucose as a delta check analyte and which is less than half of the total. 17 When choosing analytes that should undergo delta check, analytes with frequent measurement and high individuality are likely to have advantage in detecting specimen misidentification. Index of individuality is the ratio of within-subject biological variation to between-subject biological variation, and when the index of individuality is below 0.60, the analyte is considered as having high individuality. 15 Taking this into consideration, a comprehensive review of the adequacy of the delta check rule using RCV as F I G U R E 2 Comparison of the delta check detection rate using RCV with 99% CI, RCV with 99% CI excluding within-RI pairs, and RCV with 99% CI multiplied by twice the upper normal limit in inpatients (A) and outpatients (B). CI, confidence interval; RCV, reference change value; RI, reference interval well as which analyte is appropriate for laboratory error detection is required. In